Details
| Original language | English |
|---|---|
| Title of host publication | UNCECOMP 2017 |
| Subtitle of host publication | Proceedings of the 2nd International Conference on Uncertainty Quantification in Computational Sciences and Engineering |
| Editors | George Stefanou, M. Papadrakakis, Vissarion Papadopoulos |
| Pages | 154-164 |
| Number of pages | 11 |
| ISBN (electronic) | 9786188284449 |
| Publication status | Published - 2017 |
| Event | 2nd International Conference on Uncertainty Quantification in Computational Sciences and Engineering, UNCECOMP 2017 - Rhodes Island, Greece Duration: 15 Jun 2017 → 17 Jun 2017 |
Abstract
History Matching (HM) is a form of model calibration suitable for high-dimensional and computationally expensive numerical models. It sequentially cuts down the input space to find the non-implausible domain that provides a reasonable match between the output and experimental data. The non-implausible domain can be orders of magnitude smaller than the original input space and it can have a complex topology. This leads to one of the most challenging open problems in implementing HM, namely, the efficient generation of samples in the non-implausible set. Previous work has shown that Subset Simulation can be used to solve this problem. Unlike Direct Monte Carlo, Subset Simulation progressively decomposes a rare event (here is the non-implausible set), which has very small failure probabilities, into sequential less rare nested events. The original Subset Simulation uses a Modified Metropolis algorithm to generate the conditional samples that belong to intermediate less rare failure events. Generating samples moving forwards to the target space is the heart for Subset Simulation. This work considers different sampling strategies to generate samples and compares their performance in the context of expensive model calibration. A numerical example is provided to show the potential of HM using different Subset Simulation sampling schemes.
Keywords
- Bayesian emulation, History matching, Non-implausibility, Rare event simulation, Subset simulation
ASJC Scopus subject areas
- Computer Science(all)
- Computational Theory and Mathematics
- Computer Science(all)
- Computer Science Applications
- Mathematics(all)
- Theoretical Computer Science
Cite this
- Standard
- Harvard
- Apa
- Vancouver
- BibTeX
- RIS
UNCECOMP 2017: Proceedings of the 2nd International Conference on Uncertainty Quantification in Computational Sciences and Engineering. ed. / George Stefanou; M. Papadrakakis; Vissarion Papadopoulos. 2017. p. 154-164.
Research output: Chapter in book/report/conference proceeding › Conference contribution › Research › peer review
}
TY - GEN
T1 - Sampling schemes for history matching using subset simulation
AU - Gong, Z.
AU - DiazDelaO, F. A.
AU - Beer, M.
N1 - Publisher Copyright: © 2017 The Authors. Published by Eccomas Proceedia. Copyright: Copyright 2018 Elsevier B.V., All rights reserved.
PY - 2017
Y1 - 2017
N2 - History Matching (HM) is a form of model calibration suitable for high-dimensional and computationally expensive numerical models. It sequentially cuts down the input space to find the non-implausible domain that provides a reasonable match between the output and experimental data. The non-implausible domain can be orders of magnitude smaller than the original input space and it can have a complex topology. This leads to one of the most challenging open problems in implementing HM, namely, the efficient generation of samples in the non-implausible set. Previous work has shown that Subset Simulation can be used to solve this problem. Unlike Direct Monte Carlo, Subset Simulation progressively decomposes a rare event (here is the non-implausible set), which has very small failure probabilities, into sequential less rare nested events. The original Subset Simulation uses a Modified Metropolis algorithm to generate the conditional samples that belong to intermediate less rare failure events. Generating samples moving forwards to the target space is the heart for Subset Simulation. This work considers different sampling strategies to generate samples and compares their performance in the context of expensive model calibration. A numerical example is provided to show the potential of HM using different Subset Simulation sampling schemes.
AB - History Matching (HM) is a form of model calibration suitable for high-dimensional and computationally expensive numerical models. It sequentially cuts down the input space to find the non-implausible domain that provides a reasonable match between the output and experimental data. The non-implausible domain can be orders of magnitude smaller than the original input space and it can have a complex topology. This leads to one of the most challenging open problems in implementing HM, namely, the efficient generation of samples in the non-implausible set. Previous work has shown that Subset Simulation can be used to solve this problem. Unlike Direct Monte Carlo, Subset Simulation progressively decomposes a rare event (here is the non-implausible set), which has very small failure probabilities, into sequential less rare nested events. The original Subset Simulation uses a Modified Metropolis algorithm to generate the conditional samples that belong to intermediate less rare failure events. Generating samples moving forwards to the target space is the heart for Subset Simulation. This work considers different sampling strategies to generate samples and compares their performance in the context of expensive model calibration. A numerical example is provided to show the potential of HM using different Subset Simulation sampling schemes.
KW - Bayesian emulation
KW - History matching
KW - Non-implausibility
KW - Rare event simulation
KW - Subset simulation
UR - http://www.scopus.com/inward/record.url?scp=85043463649&partnerID=8YFLogxK
U2 - 10.7712/120217.5359.16948
DO - 10.7712/120217.5359.16948
M3 - Conference contribution
AN - SCOPUS:85043463649
SP - 154
EP - 164
BT - UNCECOMP 2017
A2 - Stefanou, George
A2 - Papadrakakis, M.
A2 - Papadopoulos, Vissarion
T2 - 2nd International Conference on Uncertainty Quantification in Computational Sciences and Engineering, UNCECOMP 2017
Y2 - 15 June 2017 through 17 June 2017
ER -